Implement the ConvTranspose operation

Author: mayjs

wonnx/src/compiler.rs

@@ -135,6 +135,11 @@ lazy_static! {
             include_str!("../templates/endomorphism/broadcast.wgsl"),
         )
         .unwrap();
+        tera.add_raw_template(
+            "unpool/convtranspose.wgsl",
+            include_str!("../templates/unpool/convtranspose.wgsl"),
+        )
+        .unwrap();
         tera
     };
 }
@@ -1384,6 +1389,104 @@ pub fn compile(
                 threads: (ceil(output_lengths[0], 256) as _, 1, 1),
             }
         }
+        "ConvTranspose" => {
+            /* Inputs:
+             *      1. X Input (N x C x H x W; Batch Size x Channels x Height x Weight)
+             *      2. Kernel
+             *      3. Bias
+             */
+            log::debug!("{:?}", input_shapes);
+
+            if input_shapes[0].rank() != 4 {
+                /* FIXME: We don't handle non-2D input for now */
+                return Err(CompileError::InvalidInputShape {
+                    input_index: 0,
+                    input_shape: input_shapes[0].clone(),
+                });
+            }
+
+            /* Step 1: Get the input dimensions */
+            let input_height = input_shapes[0].dims[2] as i64;
+            let input_width = input_shapes[0].dims[3] as i64;
+
+            /* Step 2: Read attributes */
+            /* TODO: auto_pad */
+            let dilations = node.get_attribute_value("dilations", Some(vec![1, 1]))?;
+
+            let group = node.get_attribute_value("group", Some(1))?;
+            if group != 1 {
+                return Err(CompileError::InvalidAttributeValue {
+                    attribute: "group".into(),
+                    value: group.to_string(),
+                    opset_version,
+                });
+            }
+
+            let inferred_kernel_shape = input_shapes[1]
+                .dims
+                .iter()
+                .skip(2)
+                .map(|&x| x as i64)
+                .collect::<Vec<_>>();
+
+            let kernel_shape =
+                node.get_attribute_value("kernel_shape", Some(inferred_kernel_shape.clone()))?;
+            if inferred_kernel_shape != kernel_shape {
+                log::error!("Inferred kernel shape: {:?}", inferred_kernel_shape);
+                return Err(CompileError::InvalidAttributeValue {
+                    attribute: "kernel_shape".to_string(),
+                    value: format!("{:?}", kernel_shape),
+                    opset_version,
+                });
+            }
+
+            let output_padding = node.get_attribute_value("output_padding", Some(vec![0, 0]))?;
+
+            let pads = node.get_attribute_value("pads", Some(vec![0, 0, 0, 0]))?;
+
+            let strides = node.get_attribute_value("strides", Some(vec![1, 1]))?;
+
+            context.insert("stride", &strides);
+
+            let output_height = strides[0] * (input_height - 1)
+                + output_padding[0]
+                + ((kernel_shape[0] - 1) * dilations[0] + 1)
+                - pads[0]
+                - pads[2];
+            let output_width = strides[1] * (input_width - 1)
+                + output_padding[1]
+                + ((kernel_shape[1] - 1) * dilations[1] + 1)
+                - pads[1]
+                - pads[3];
+
+            log::debug!(
+                "Calculated output size: {:?}x{:?}",
+                output_width,
+                output_height
+            );
+
+            let (x_threads, workgroup_size_x) = workgroup_size(
+                output_lengths[0],
+                MAX_COMPUTE_WORKGROUPS_PER_DIMENSION,
+                MAX_WORKGROUP_SIZE_X,
+            )?;
+            context.insert("workgroup_size_x", &workgroup_size_x);
+
+            let scalar_type = agreed_type(input_shapes, output_shapes)?;
+
+            if scalar_type.is_float() {
+                NodeTemplate {
+                    scalar_type,
+                    template: "unpool/convtranspose.wgsl",
+                    threads: (x_threads, 1, 1),
+                }
+            } else {
+                return Err(CompileError::UnimplementedVariant {
+                    variant: "Non-Float".into(),
+                    op: "ConvTranspose".into(),
+                });
+            }
+        }
         op => return Err(CompileError::UnimplementedOp(op.to_string())),
     };
 

wonnx/templates/unpool/convtranspose.wgsl

@@ -0,0 +1,70 @@
+{%- include "structs.wgsl" -%}
+
+@group(0) @binding(0)
+var<storage, read> input_tensor: Array;
+
+@group(0) @binding(1)
+var<storage, read> input_kernel_weights: Array;
+
+{% if i_lens | length == 3 -%}
+    @group(0) @binding(2)
+    var<storage, read> input_bias: Array;
+
+    @group(0) @binding(3)
+    var<storage, read_write> output_0: Array;
+{%- else -%}
+    @group(0) @binding(2)
+    var<storage, read_write> output_0: Array;
+{%- endif %}
+
+@compute @workgroup_size({{ workgroup_size_x }}, 1, 1)
+fn main(@builtin(global_invocation_id) global_id: vec3<u32>) {
+    let output_idx = global_id.x;
+
+    if (output_idx < {{ o_lens[0] }}u) {
+        let batch = output_idx / {{ o_chunks[0][0] }}u;
+        var rest = output_idx % {{ o_chunks[0][0] }}u;
+
+        let channel = rest / {{ o_chunks[0][1] }}u;
+        rest = rest % {{ o_chunks[0][1] }}u;
+
+        let y = rest / {{ o_chunks[0][2] }}u;
+        let x = rest % {{ o_chunks[0][2] }}u;
+
+        let sample_root_index = batch * {{ i_chunks[0][0] }}u;
+
+        /* Kernel Size: C x M/groups x kH x kW */
+        /* -> For each ouput channel, we need to calculate the convolution over each input channel and sum all of that up! */
+        /* C = In Channels */
+        /* M = Out Channels */
+
+        /* Kernel base coordinate limits */
+        let min_in_y = select(0u, (y - {{ i_shape[1][2] }}u) / {{ stride[0] }}u, y > {{ i_shape[1][2] }}u);
+        let max_in_y = select({{ i_shape[0][2] }}u - 1u, y / {{ stride[0] }}u, y / {{ stride[0] }}u < {{ i_shape[0][3] }}u);
+        let min_in_x = select(0u, (x - {{ i_shape[1][3] }}u) / {{ stride[1] }}u, x > {{ i_shape[1][3] }}u);
+        let max_in_x = select({{ i_shape[0][3] }}u - 1u, x / {{ stride[1] }}u, x / {{ stride[1] }}u < {{ i_shape[0][3] }}u);
+
+        var result: Scalar = Scalar();
+
+        for(var ichannel: u32 = 0u; ichannel < {{ i_shape[0][1] }}u; ichannel = ichannel + 1u) {
+            let base_index = sample_root_index + ichannel * {{ i_chunks[0][1] }}u;
+            let base_kernel_index = {{ i_chunks[1][0] }}u * ichannel + channel * {{ i_chunks[1][1] }}u; /* Base kernel address to apply kH and kW to */
+
+            for(var in_y: u32 = min_in_y; in_y <= max_in_y; in_y = in_y + 1u) {
+                for(var in_x: u32 = min_in_x; in_x <= max_in_x; in_x = in_x + 1u) {
+                    let kernel_y = y - (in_y * {{ stride[0] }}u);
+                    let kernel_x = x - (in_x * {{ stride[1] }}u);
+
+                    if(kernel_y < {{ i_shape[1][2] }}u && kernel_x < {{ i_shape[1][3] }}u) {
+                        result = result + (input_tensor.data[base_index + (in_y * {{ i_chunks[0][2] }}u) + in_x]
+                                           * input_kernel_weights.data[base_kernel_index + kernel_y * {{ i_chunks[1][2] }}u + kernel_x]);
+                    }
+                }
+            }
+        }
+        {% if i_lens | length == 3 -%}
+            result = result + input_bias.data[channel];
+        {%- endif %}
+        output_0.data[output_idx] = result;
+    }
+}

wonnx/tests/convtranspose.rs

@@ -0,0 +1,93 @@
+use std::collections::HashMap;
+use wonnx::utils::{attribute, graph, initializer, model, node, tensor, OutputTensor};
+mod common;
+
+#[test]
+fn convtranspose_default() {
+    let data: Vec<f32> = vec![0.0, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0];
+    let input_shape = vec![1, 1, 3, 3];
+
+    let data_w = vec![
+        1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0,
+    ];
+    let kernel_shape = vec![1, 2, 3, 3];
+
+    let output_shape = vec![1, 2, 5, 5];
+
+    let input_data = HashMap::from([("X".to_string(), data.as_slice().into())]);
+
+    let convtranpose_model = model(graph(
+        vec![tensor("X", &input_shape)],
+        vec![tensor("Y", &output_shape)],
+        vec![],
+        vec![initializer("W", data_w, kernel_shape)],
+        vec![node(
+            vec!["X", "W"],
+            vec!["Y"],
+            "convtranspose",
+            "ConvTranspose",
+            vec![attribute("kernel_shape", vec![3, 3])],
+        )],
+    ));
+
+    let session = pollster::block_on(wonnx::Session::from_model(convtranpose_model))
+        .expect("Session did not create");
+    let result = pollster::block_on(session.run(&input_data)).unwrap();
+
+    assert_eq!(
+        result["Y"],
+        OutputTensor::F32(vec![
+            0.0, 1.0, 3.0, 3.0, 2.0, 3.0, 8.0, 15.0, 12.0, 7.0, 9.0, 21.0, 36.0, 27.0, 15.0, 9.0,
+            20.0, 33.0, 24.0, 13.0, 6.0, 13.0, 21.0, 15.0, 8.0, 0.0, 1.0, 3.0, 3.0, 2.0, 3.0, 8.0,
+            15.0, 12.0, 7.0, 9.0, 21.0, 36.0, 27.0, 15.0, 9.0, 20.0, 33.0, 24.0, 13.0, 6.0, 13.0,
+            21.0, 15.0, 8.0,
+        ])
+    );
+}
+
+#[test]
+fn convtranspose_strides() {
+    let data = vec![0.0, 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0]; // (1, 1, 3, 3)
+    let input_shape = vec![1, 1, 3, 3];
+
+    let data_w = vec![
+        1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0,
+    ];
+    let kernel_shape = vec![1, 2, 3, 3];
+
+    let output_data = vec![
+        0.0, 0.0, 1.0, 1.0, 3.0, 2.0, 2.0, 0.0, 0.0, 0.0, 1.0, 1.0, 3.0, 2.0, 2.0, 0.0, 0.0, 0.0,
+        1.0, 1.0, 3.0, 2.0, 2.0, 0.0, 3.0, 3.0, 7.0, 4.0, 9.0, 5.0, 5.0, 0.0, 3.0, 3.0, 7.0, 4.0,
+        9.0, 5.0, 5.0, 0.0, 3.0, 3.0, 7.0, 4.0, 9.0, 5.0, 5.0, 0.0, 6.0, 6.0, 13.0, 7.0, 15.0, 8.0,
+        8.0, 0.0, 6.0, 6.0, 13.0, 7.0, 15.0, 8.0, 8.0, 0.0, 6.0, 6.0, 13.0, 7.0, 15.0, 8.0, 8.0,
+        0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0, 3.0, 2.0, 2.0, 0.0, 0.0,
+        0.0, 1.0, 1.0, 3.0, 2.0, 2.0, 0.0, 0.0, 0.0, 1.0, 1.0, 3.0, 2.0, 2.0, 0.0, 3.0, 3.0, 7.0,
+        4.0, 9.0, 5.0, 5.0, 0.0, 3.0, 3.0, 7.0, 4.0, 9.0, 5.0, 5.0, 0.0, 3.0, 3.0, 7.0, 4.0, 9.0,
+        5.0, 5.0, 0.0, 6.0, 6.0, 13.0, 7.0, 15.0, 8.0, 8.0, 0.0, 6.0, 6.0, 13.0, 7.0, 15.0, 8.0,
+        8.0, 0.0, 6.0, 6.0, 13.0, 7.0, 15.0, 8.0, 8.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0,
+    ];
+    let output_shape = vec![1, 2, 10, 8];
+
+    let convtranpose_model = model(graph(
+        vec![tensor("X", &input_shape)],
+        vec![tensor("Y", &output_shape)],
+        vec![],
+        vec![initializer("W", data_w, kernel_shape)],
+        vec![node(
+            vec!["X", "W"],
+            vec!["Y"],
+            "convtranspose",
+            "ConvTranspose",
+            vec![
+                attribute("kernel_shape", vec![3, 3]),
+                attribute("strides", vec![3, 2]),
+            ],
+        )],
+    ));
+
+    let input_data = HashMap::from([("X".to_string(), data.as_slice().into())]);
+    let session = pollster::block_on(wonnx::Session::from_model(convtranpose_model))
+        .expect("Session did not create");
+    let result = pollster::block_on(session.run(&input_data)).unwrap();
+    assert_eq!(result["Y"], OutputTensor::F32(output_data));
+}